Re: slow commits with heavy temp table usage in 8.4.0

"Todd A. Cook" <tcook(at)blackducksoftware(dot)com> writes:
> I've noticed that on 8.4.0, commits can take a long time when a temp table is repeatedly
> filled and truncated within a loop. A very contrived example is

Hmm. I tweaked the function to allow varying the number of truncates:

regression=# begin;
BEGIN
Time: 1.037 ms
regression=# select commit_test_with_truncations(10000) ;
commit_test_with_truncations
------------------------------

(1 row)

Time: 9466.060 ms
regression=# commit;
COMMIT
Time: 1095.946 ms
regression=# begin;
BEGIN
Time: 1.002 ms
regression=# select commit_test_with_truncations(30000) ;
commit_test_with_truncations
------------------------------

(1 row)

Time: 93492.874 ms
regression=# commit;
COMMIT
Time: 3184.248 ms

The commit time doesn't seem tremendously out of line, but it looks
like there's something O(N^2)-ish in the function execution. Do
you see a similar pattern? With so many temp files there could well
be some blame on the kernel side. (This is a Fedora 10 box.)

regards, tom lane